Personalization of a user interface using machine learning

Abstract

A unique implementation of a machine learning application for suggesting actions for a user to undertake is described herein. The application transforms a history of user behavior into a set of models that represent user actions given a set of parameters. These models are then used to suggest that users in a payments or banking environment take certain actions based on their history. The models are created using the DensiCube, random forest or k-means algorithms.

The invention claimed is: 
     
       1. A method for automatically suggesting actions to a user in a user interface comprising:
 receiving a set of input parameters, wherein the set of input parameters include information regarding time and user task; 
 accessing a list of possible actions; 
 filtering the list of possible actions to remove the actions that are not available to the user; 
 looping through the filtered list of possible actions until the filtered list is processed,
 executing a machine learning model of user behavior on each possible action in the filtered list with the set of input parameters to obtain a machine learning score; 
 storing the machine learning score with the possible action in the filtered list; 
 
 once the filtered list of possible actions is processed, sorting the filtered list of possible actions by the machine learning score; 
 selecting the possible actions from the filtered list with high machine learning scores; and 
 offering the user options to perform the selected possible actions; 
 wherein the machine learning model is built by iterating through possible rule sets to identify the rule set with a best quality score using a data set of previous user behavior wherein the data set of previous user behavior utilizes the data of multiple users up to a threshold of data items for a specific user and then switches to solely using data from the specific user. 
 
     
     
       2. The method of  claim 1  wherein the machine learning model is built using a DensiCube machine learning algorithm. 
     
     
       3. The method of  claim 2  wherein the machine learning model is built using a distributed DensiCube machine learning algorithm. 
     
     
       4. The method of  claim 1  wherein the machine learning model is built using a K-means machine learning algorithm. 
     
     
       5. The method of  claim 1  wherein the machine learning model is built using a random forest machine learning algorithm. 
     
     
       6. The method of  claim 1  wherein the data set of previous user behavior utilizes a graduated combination of data of the multiple users combined with the data from a specific user weighted by a factor based upon a number of data items for the specific user. 
     
     
       7. The method of  claim 1  further comprising, before the looping through the filtered list of possible actions,
 accessing a list of possible situations; 
 looping through the list of possible situations until the list of possible situations is processed,
 executing a situations machine learning model of user behavior on each possible situation with the input parameters to obtain a situations machine learning score; and 
 storing the situations machine learning score with the possible situation; 
 
 once the list of possible situations is processed, sorting the list of possible situations by the score; and 
 selecting the possible situations with high scores. 
 
     
     
       8. The method of  claim 1  further comprising automatically undertaking the action. 
     
     
       9. The method of  claim 1  wherein the input parameters also include location information. 
     
     
       10. A system for that automatically suggests actions to a user in a user interface comprising:
 a special purpose server; 
 a data storage device electrically connected to the special purpose server, where the data storage device holds a history of user behavior, a list of possible actions, and models of the user behavior; 
 an internet connected to the special purpose server; 
 a computing device connected to the special purpose server through the internet;
 wherein the user uses the computing device to log into an application, the application sends input parameters to the special purpose server, wherein the input parameters include information regarding time and user task, the special purpose server accesses the list of possible actions and filters the list of possible actions to remove the actions that are not available to the user, wherein the special purpose server loops through the filtered list of possible actions until the list is processed, while the special purpose server executes a machine learning model on each possible action in the filtered list with the input parameters to obtain a machine learning score and stores the machine learning score with the possible action in the filtered list; and 
 once the special purpose server processes the filtered list of possible actions, the special purpose server sorts the filtered list of possible actions by the score, selects the possible actions in the filtered list with high scores, and offers the user options to perform the selected possible actions; 
 
 wherein the machine learning model is built by the special purpose server through iterations of possible rule sets to identify the rule set with a best quality score using a data set of previous user behavior; 
 wherein the data set of previous user behavior utilizes the data of multiple users up to a threshold of data items for a specific user and then switches to solely using data from the specific user. 
 
     
     
       11. The system of  claim 10  wherein the machine learning model is built with a DensiCube machine learning algorithm. 
     
     
       12. The system of  claim 11  wherein the machine learning model is built with a distributed DensiCube machine learning algorithm. 
     
     
       13. The system of  claim 10  wherein the machine learning model is built with a K-means machine learning algorithm. 
     
     
       14. The system of  claim 10  wherein the machine learning model is built with a random forest machine learning algorithm. 
     
     
       15. The system of  claim 10  wherein the data set of previous user behavior utilizes a graduated combination of the data of the multiple users combined with the data from a specific user weighted by a factor based upon a number of data items for the specific user. 
     
     
       16. The system of  claim 10  wherein the input parameters also include location information.